在現今的研究報告中,關於BGA(Ball Grid Array)表面瑕疵的偵測,業者方面仍希望能更提高速度或正確性。而在傳統上,BGA檢測大都使用灰階影像來進行。但由於焊墊、導線及基板在影像中所呈現的灰階值會差異不大,所以當臨界值設定不良或對比度相差過小時,即會產生無法分割出目標物的情形。 本研究主要目的有三。第一、本研究希望針對BGA之表面均勻度以及外型特徵,建立分類模式與進行瑕疵辨識,不進行樣本比對。第二、應用彩色機器視覺的系統來進行分割檢測。第三、改善整個系統檢測的速度與效能。 傳統的影像強化流程是先選擇適合的色彩模型,選定之後以此模型進行強化,如直方均衡強化法;本研究提出使用Gamma修正法來取代傳統的影像強化流程,以期能有更好的影像強化效果。當使用Gamma修正法處理之後,發現因為Gamma修正法針對三色頻全面修正,所以由Gamma修正法所進行的強化可以明顯地分開高低對比之間的影像;而於分割時使用Gamma修正法配合R色頻進行分割可得到將背景與前景分開的分割結果,平均改善對比值達52.09%。最後本研究以外型和不均勻瑕疵指標互相配合,對BGA之錫球與線路進行檢測,發現可將96.43%的瑕疵分判出來;而在速度的改善方面,使用傳統的強化方法對一640*480像素大小的圖片進行完整的瑕疵偵測需1秒,而使用Gamma修正法來強化的結果是在0.3秒左右可以完成,針對業界需要大量檢測時,可以節省相當多的時間。 本論文經由最後實驗結果可得到由Gamma修正法所進行的強化可以使用彩色機器視覺分開高低對比之間的影像,而速度亦能滿足要求;以外型和不均勻瑕疵指標互相配合,可將瑕疵分判出來而不使用樣板比對。因此可以提供給業界一快速的檢測方法並克服傳統強化分割上的問題。
In the current manufacturing environment, a company still needs to have a faster and a more accurate ways to inspect a Ball Grid Array (BGA) surface defects. Traditionally, the BGA inspection was using gray-level images. However, the background, conduct paths and pads have very similar gray-levels that cannot easy be distinguished. The objectives of the research are: (1) Use some shape and uniformity features without making pattern matching for detecting BGA surface defects. (2) Use color image information instead of gray scales for the inspection. (3) Improve the speed and effectiveness of the inspection system. The traditional process of the image enhancement is to select a suitable color mode and then to proceed on the enhancement. The research proposed a method that to use the gamma correction method to replace the tradition process for image enhancement with the expectations of having better results and faster speeds. Because gamma correction corrects the three color bands (i.e., RGB), it could better separate the image between the high and low contrasts. And it could get the better results in dividing the image into background and foreground by using the Gamma correction and the R color band. As a result, the proposed method can improve the contrast value about 52.09%. Finally, the research uses the eigenvalues of the shape and uniformity to detect the defects. It could find almost all the defects. In using the traditionally enhancement method with a 640* 480 pixels image to do a completely defects detecting needs 1 sec. However, to use the proposed gamma correction method to do the same, needs only 0.3 sec. This research demonstrates the effectiveness of using gamma correction method for separating an image background from its foreground. And the developed method could detect the BGA surface defects without using pattern matching technique that required extensive alignment in both hardware and software.